Rackjure | Hacker News Reader

Rackjure

docs.racket-lang.org

147 pointstosh29 comments

I guess I'm alone on this, but I want as little special syntax as possible, it's the primary thing I like about Scheme. ' ` , # <- that's enough, let's stop there. Everything else should be in an s-expression.

Part of the joy of Racket is that you can mix and match :-)

In some cases, it allows people who like some of the Clojure syntax to still write you libraries. In other cases it allows you to move a prototype from normal Racket to Typed Racket in a piecemeal way.

In general I like the purity of scheme as well, but sometimes that little bit of syntactic sugar goes a long way. Non-Racket Schemers use macros too, and Racket’s are second to none. Use what you want and avoid the rest!

Well, I suggest not using anything but scheme.

Downvoted for facile snark.

It's not special syntax just because the identifier is some symbols. If you don't like the semantics, don't use it; but it's not correct to accuse it of being "special syntax".

Edit: oh, and if you _do_ like the semantics, but hate the weird name, just give it a new name that makes sense to you.

(that-macro-from-clojure ....)

Clojure does have a number of sigils and special syntax, but they all seem to have s-expression alternatives. e.g. a dict can be created with {:a 1, :b 2} or (hash-map :a 1, :b 2); dereferencing an atom can be done with @a or (deref a) and so on.

that's a pretty arbitrary stopping point, don't you think?

it's not like clojure added that much to the list: [], {} and @

but it is lightyears ahead in convenience

In my opinion one of the primary characteristics of Clojure is the implementation of the collection types. Without that it's not going to be an implementation of Clojure to me.

Yep. I'd say persistent collections and (lazy) sequences. And great concurrency stuff (agents, atoms etc.)

This. Without persistent collections, any lisp that looks like Clojure is only half way there.

On the other hand the author is pretty explicit that this is a limited borrow from Clojure, and that when in conflict, he’ll do the more Rackety thing.

This got me curious about whether persistent data structures are available in Racket. It seems like they are.

https://docs.racket-lang.org/functional-data-structures/inde...

Being fundamental to clojure means everyone uses those data structures. That matters quite a lot when you aren't writing everything from scratch.

There is also immutable hash-tables available: http://docs.racket-lang.org/reference/hashtables.html

(Just to be clear: there is of course also mutable hash tables available).

Shameless plug: I'm writing similar thing as a hobby sandbox project: https://github.com/kovrik/scheme-in-kotlin

It started as a Scheme R5RS implementation in Java, then I translated everything to Kotlin. Finally, decided to make it a hybrid (abomination?) of Scheme and Racket and Clojure.

Still in development, don't know how to implement macros properly. Also, it's only an interpreter for now, there is no compiler (for the same reason: don't know yet how to write a compiler for Kotlin. Use ASM?)

Anyway, it is a great and enlightening experience and I would recommend everyone to write your own lang.

Why are there two variations for ~> and ~>> ? It looks like it limits you to using only functions that take parameters in the same position.

Why not have one position-independent version? Something where you can specify at every level which position you want to use, like this:

    (~>>> #"foobar"
          (~> bytes-length)
          (~>> number->string 16)
          (~> string->bytes/utf-8))

But that still only limits you to first and last position. Why not use a positional special form like JavaScript has with __dirname called __var?

    (~>>> #"foobar"
          (__var bytes-length)
          (number->string 16 __var)
          (__var string->bytes/utf-8))

This way you can use it no matter what functions you want. But I don't write my code in Racket so maybe this is already a solved problem.

Clojure has that, it's called as->, and you specify a symbol to represent the previous result in each subform.

But the first/last variants, -> and ->>, are still useful because they fit a subtle design decision in Clojure. Fns that operate on objects/scalars tend to have the object modified as the very first parameter, while collection-operating fns tend to have the collection last.

In clojure, you can do this with as->, but there are reasons to be careful. Stuart Sierra has a great (brief) article about it[0]. The long and short of it is that it's really easy to change the "type" of an expression as it goes through your pipeline in unintuitive ways. This can cause surprises later down the line, so it's probably better to refactor the functions you are calling so that they play nicely with thread-first or with thread-last.

[0]https://stuartsierra.com/2018/07/15/clojure-donts-thread-as

Did you mean Stuart Sierra? :-)

Whoops! Clojure is the only community I know with a confusion of Stuarts.

I feel the same way. I also tend to forget which macro inserts argument at front and which one at back. I tend to use threading[1] package for macros which does what you are proposing:

[1] https://docs.racket-lang.org/threading/index.html#%28form._%...

These days that's what rackjure re-provides: http://docs.racket-lang.org/rackjure/index.html#(mod-path._r...

Aside from `as->` which mentioned by others, you can nest thread-last into thread-first if you need to and it behaves appropriately. (You can't go the other way for obvious reasons but it works when you introduce a collection.)

People have already described the form that already exists in Clojure, but it's worth mentioning that this kind of macro is called an anaphoric macro and is featured in Let Over Lambda (https://letoverlambda.com/index.cl/guest/chap6.html) and On Lisp. They tend not to be viewed as idiomatic Clojure. I would assume they are less used in Schemes since Scheme macros are hygienic by design and anaphoric macros depend on variable capture. That said I'm not a Schemer and may be wrong. These kind of macros are very much part of the Lisp heritage though.

I would have called it Clacket. Cool project nonetheless.

Portmanteaus for naming new technology should be banned :P

> portechteaus

FTFY

Would love to see an Arcket.

I think you can use Anarki in exactly that way:

https://github.com/arclanguage/anarki#racket-interop

Reply on news.ycombinator.com